Added: Generate image to video

.env

@@ -1,2 +1,5 @@
 SERVER_ADDRESS=https://gosign-de-image-to-video.hf.space
-WS_ADDRESS=wss://gosign-de-image-to-video.hf.space/ws
+WS_ADDRESS=wss://gosign-de-image-to-video.hf.space/ws
+
+# SERVER_ADDRESS=http://127.0.0.1:8188/
+# WS_ADDRESS=ws://127.0.0.1:8188/ws

app.py

@@ -12,6 +12,8 @@ from dotenv import load_dotenv
 from flask import Flask, request, jsonify, render_template, send_file, send_from_directory
 from PIL import Image
 from werkzeug.utils import secure_filename
+import urllib.parse
+import urllib.request
 
 # Load environment variables from the .env file
 load_dotenv()
@@ -42,7 +44,7 @@ def get_image(filename, subfolder, image_type, token):
 
 
 def get_images(ws, workflow, token):
-    prompt_id = queue_prompt(workflow, token)['prompt_id']
+    prompt_id = queue_prompt(workflow, token)
     output_images = {}
 
     while True:
@@ -66,13 +68,6 @@ def get_images(ws, workflow, token):
 
     return output_images
 
-
-def fetch_video(video_data, token):
-    video_url = f"{server_address}/download?file={video_data['filename']}"
-    req = urllib.request.Request(video_url)
-    req.add_header("Authorization", f"Bearer {token}")
-    return urllib.request.urlopen(req).read()
-
 # Default route for home welcome
 @app.route('/')
 def home():
@@ -226,18 +221,47 @@ def get_history(prompt_id, token):
     return make_request(f"{server_address}/history/{prompt_id}", headers=headers)
 
 def get_video_data(filename, subfolder, token):
-    url_values = {'filename': filename, 'subfolder': subfolder}
+    """
+    Retrieve a video from the server using filename, subfolder, and token.
+    """
+    # Handle empty subfolder case gracefully
+    subfolder = subfolder or ''  # Default to empty string if None
+
+    # Construct query parameters
+    # url_values = {
+    #     'filename': filename,
+    #     'subfolder': subfolder,
+    #     'type': 'video'
+    # }
+
+     # Construct query parameters
+    url_values = {
+        'filename': filename
+    }
+
+    # Build the URL with encoded query parameters
     url = f"{server_address}/view?{urllib.parse.urlencode(url_values)}"
+    print(f"Requesting URL: {url}")
+
+    # Prepare the request with authorization token
     req = urllib.request.Request(url)
     req.add_header("Authorization", f"Bearer {token}")
 
     try:
-        return urllib.request.urlopen(req).read()  # Return video content
+        # Fetch and return the video data
+        response = urllib.request.urlopen(req)
+        return response.read()
+
     except urllib.error.HTTPError as e:
         print(f"HTTP Error: {e.code} - {e.reason}")
-        print(e.read())
+        print(e.read().decode())  # Decode error message for readability
         raise
 
+    except urllib.error.URLError as e:
+        print(f"URL Error: {e.reason}")
+        raise
+
+
 # Route: Image to Video
 @app.route('/image_to_video', methods=['POST'])
 def image_to_video():
@@ -259,19 +283,31 @@ def image_to_video():
     base64_image = data.get('base64_image')
 
     if image_file:
-        # Validate and save the uploaded image
+        # Check if the file has an allowed extension
         if not allowed_file(image_file.filename):
             return jsonify({'error': 'Unsupported image format'}), 400
+
+        # Secure the filename
         filename = secure_filename(image_file.filename)
-        image_path = f"/tmp/{uuid.uuid4()}_{filename}"
 
+        # Generate a unique path for the image
+        unique_filename = f"{uuid.uuid4()}_{filename}"
+        image_path = os.path.join('static', unique_filename)
 
+        # Ensure the 'static' directory exists
+        os.makedirs('static', exist_ok=True)
+
+        # Save the image to the static directory
         image_file.save(image_path)
+
+        # Construct the public URL to access the image
+        image_url = f"https://gosign-de-comfyui-api.hf.space/{image_path}"
+
     elif base64_image:
         # Save base64 image
         try:
-            image_path = save_base64_image(base64_image)
-            # return jsonify({'image_path': image_path})
+            image_url = save_base64_image(base64_image)
+            # return jsonify({'image_url': image_url})
 
         except Exception as e:
             return jsonify({'error': f'Invalid base64 image data: {str(e)}'}), 400
@@ -286,7 +322,7 @@ def image_to_video():
 
     # Modify workflow with inputs
     workflow["30"]["inputs"]["prompt"] = text_prompt
-    # workflow["73"]["inputs"]["url"] = image_path
+    workflow["73"]["inputs"]["url"] = image_url
     workflow["31"]["inputs"]["prompt"] = "Low quality, watermark, strange motion"
     workflow["57"]["inputs"]["seed"] = random.randint(100000000000000, 999999999999999)
 
@@ -308,29 +344,45 @@ def image_to_video():
         if message.get('type') == 'executing' and message['data']['node'] is None and message['data']['prompt_id'] == prompt_id:
             break
 
-    # Fetch video details
+   # Fetch the complete history for the provided prompt_id
     history = get_history(prompt_id, token).get(prompt_id, {})
-    video_data = None
-    for node_id, node_output in history.get('outputs', {}).items():
-        if 'videos' in node_output:
-            video = node_output['videos'][0]
-            video_data = get_video_data(video['filename'], video['subfolder'], token)
+    print(f"History for prompt {prompt_id}: {history}")
+
+    video_data = None  # Initialize video data
 
+    # Loop through history outputs to find video or gif data
+    for node_id, node_output in history.get('outputs', {}).items():
+        if 'gifs' in node_output:
+            video = node_output['gifs'][0]  # Take the first GIF or video
+            try:
+                print(f"Fetching video: {video['filename']} from {video['subfolder']}")
+                video_data = get_video_data(video['filename'], video['subfolder'], token)
+                break  # Stop after successfully fetching the first video or GIF
+            except Exception as e:
+                print(f"Failed to retrieve video: {str(e)}")
+
+    # Check if video data was retrieved successfully
     if not video_data:
         return jsonify({'error': 'Failed to generate video'}), 500
 
-    # Save and return the generated video
-    local_video_path = f"/tmp/generated_video_{uuid.uuid4()}.mp4"
-    with open(local_video_path, 'wb') as f:
-        f.write(video_data)
+    # Save the video locally
+    # local_video_path = f"/tmp/generated_video_{uuid.uuid4()}.mp4"
+    # with open(local_video_path, 'wb') as f:
+    #     f.write(video_data)
 
-    return send_file(
+    # Return the video as an HTTP response
+    response = send_file(
         io.BytesIO(video_data),
         mimetype='video/mp4',
         as_attachment=True,
         download_name='generated_video.mp4'
     )
 
+    # Delete the saved image after sending the response
+    if image_path and os.path.exists(image_path):
+        os.remove(image_path)  # Remove the image file
+
+    return response
 
 if __name__ == '__main__':
     app.run(host='0.0.0.0', port=7860, debug=True)  # Removed 'debug=True'

sample-response.json

@@ -0,0 +1,54 @@
+{'prompt': [
+        3, '2d7683a1-f132-4f75-bb02-ced4975a745b',
+        {'1': {'inputs': {'model': 'THUDM/CogVideoX-5b-I2V', 'precision': 'bf16', 'fp8_transformer': 'disabled', 'compile': 'disabled', 'enable_sequential_cpu_offload': False
+                }, 'class_type': 'DownloadAndLoadCogVideoModel', '_meta': {'title': '(Down)load CogVideo Model'
+                }
+            }, '20': {'inputs': {'clip_name': 't5\\google_t5-v1_1-xxl_encoderonly-fp8_e4m3fn.safetensors', 'type': 'sd3'
+                }, 'class_type': 'CLIPLoader', '_meta': {'title': 'Load CLIP'
+                }
+            }, '30': {'inputs': {'prompt': 'A beautiful sunset over the mountains', 'strength': 1.0, 'force_offload': True, 'clip': ['20',
+                        0
+                    ]
+                }, 'class_type': 'CogVideoTextEncode', '_meta': {'title': 'CogVideo TextEncode'
+                }
+            }, '31': {'inputs': {'prompt': 'Low quality, watermark, strange motion', 'strength': 1.0, 'force_offload': True, 'clip': ['20',
+                        0
+                    ]
+                }, 'class_type': 'CogVideoTextEncode', '_meta': {'title': 'CogVideo TextEncode'
+                }
+            }, '37': {'inputs': {'width': 720, 'height': 480, 'upscale_method': 'lanczos', 'keep_proportion': False, 'divisible_by': 16, 'crop': 'disabled', 'image': ['73',
+                        0
+                    ]
+                }, 'class_type': 'ImageResizeKJ', '_meta': {'title': 'Resize Image'
+                }
+            }, '44': {'inputs': {'frame_rate': 12.0, 'loop_count': 0, 'filename_prefix': 'CogVideoX-I2V', 'format': 'video/h264-mp4', 'pix_fmt': 'yuv420p', 'crf': 19, 'save_metadata': True, 'pingpong': False, 'save_output': True, 'images': ['56',
+                        0
+                    ]
+                }, 'class_type': 'VHS_VideoCombine', '_meta': {'title': 'Video Combine 🎥🅥🅗🅢'
+                }
+            }, '56': {'inputs': {'enable_vae_tiling': False, 'tile_sample_min_height': 96, 'tile_sample_min_width': 96, 'tile_overlap_factor_height': 0.083, 'tile_overlap_factor_width': 0.083, 'auto_tile_size': True, 'pipeline': ['57',
+                        0
+                    ], 'samples': ['57',
+                        1
+                    ]
+                }, 'class_type': 'CogVideoDecode', '_meta': {'title': 'CogVideo Decode'
+                }
+            }, '57': {'inputs': {'height': 480, 'width': 720, 'num_frames': 49, 'steps': 5, 'cfg': 6.0, 'seed': 660821088584312, 'scheduler': 'DPM++', 'denoise_strength': 1.0, 'pipeline': ['1',
+                        0
+                    ], 'positive': ['30',
+                        0
+                    ], 'negative': ['31',
+                        0
+                    ], 'image_cond_latents': ['58',
+                        0
+                    ]
+                }, 'class_type': 'CogVideoSampler', '_meta': {'title': 'CogVideo Sampler'
+                }
+            }, '58': {'inputs': {'chunk_size': 16, 'enable_tiling': True, 'pipeline': ['1',
+                        0
+                    ], 'image': ['37',
+                        0
+                    ]
+                }, 'class_type': 'CogVideoImageEncode', '_meta': {'title': 'CogVideo ImageEncode'
+                }
+            }, '73': {'inputs': {'url': 'https: //huggingface.co/spaces/gosign-de/comfyui-api/resolve/main/images/iStock_000014226797_Small.jpg', 'cache': True}, 'class_type': 'LoadImageByUrl //Browser', '_meta': {'title': 'Load Image By URL'}}, '75': {'inputs': {'images': ['73', 0]}, 'class_type': 'PreviewImage', '_meta': {'title': 'Preview Image'}}}, {'client_id': 'bb7de022-33ad-4c01-b272-bf2b1eb05e4e'}, ['75', '44']], 'outputs': {'44': {'gifs': [{'filename': 'CogVideoX-I2V_00003.mp4', 'subfolder': '', 'type': 'output', 'format': 'video/h264-mp4', 'frame_rate': 12.0}]}, '75': {'images': [{'filename': 'ComfyUI_temp_favdf_00001_.png', 'subfolder': '', 'type': 'temp'}]}}, 'status': {'status_str': 'success', 'completed': True, 'messages': [['execution_start', {'prompt_id': '2d7683a1-f132-4f75-bb02-ced4975a745b', 'timestamp': 1729158615147}], ['execution_cached', {'nodes': ['1', '20', '30', '31', '37', '58', '73', '75'], 'prompt_id': '2d7683a1-f132-4f75-bb02-ced4975a745b', 'timestamp': 1729158615162}], ['execution_success', {'prompt_id': '2d7683a1-f132-4f75-bb02-ced4975a745b', 'timestamp': 1729158723213}]]}, 'meta': {'44': {'node_id': '44', 'display_node': '44', 'parent_node': None, 'real_node_id': '44'}, '75': {'node_id': '75', 'display_node': '75', 'parent_node': None, 'real_node_id': '75'}}}

workflows/cogvideox_image_to_video_workflow_api.json

@@ -118,7 +118,7 @@
       "height": 480,
       "width": 720,
       "num_frames": 49,
-      "steps": 10,
+      "steps": 2,
       "cfg": 6,
       "seed": 65334758276105,
       "scheduler": "DPM++",